1. Field of Invention
This invention is directed to spectrum inverter apparatus and methods that provide a full spectrum representation of the true color of an image based on input color values and a medium identifier.
2. Description of Related Art
When images are reproduced using conventional scanning processes, the resulting reproduction of an image is not an accurate reproduction of the original image. This is because the scanning devices use a limited color space. For example, spectral reflectance values, i.e. the actual color values reflected from or through the surface of the image, may be represented by a 36-dimensional spectral space while the scanner devices are limited to using a 3-dimensional color space, such as CIE Lab, RGB (red, green, blue), or the like. Thus, different image values in the 36-dimensional spectral space may be mapped to the same 3-dimensional color space values. This is known as metamerism.
Thus, due to metamerism, an image that is scanned by a scanning device does not completely represent the actual physical image. Furthermore, the reproduced image does not completely represent the original image, since many of the image values in the 36-dimensional spectral space have been lost during scanning of the image.
This invention is directed to spectrum inverter methods and apparatus. The spectrum inverter receives optical values, such as scanned input color values, from a scanning device and applies a spectrum inversion method to the optical values. The spectrum inversion method is selected based on a medium identifier that identifies the medium upon which the scanned image is formed. The resulting spectral space values are then output and may be used by an image output device to reproduce the scanned image.
In one exemplary embodiment of the spectrum inversion apparatus and methods of this invention, the spectrum inversion is implemented using a neural network that has three input layer nodes, ten hidden layer nodes, and thirty six output layer nodes. The optical values received from the scanning device are input into the input layer nodes and are inverted by the hidden and output layer nodes to output a 36-dimensional spectral space representation of the optical values. The weights of the nodes in the neural network are selected based on stored weight information corresponding to the medium identified by the medium identifier. These weights may be determined based on training the neural network using training data. Thus, a different set of weights may be used depending on the particular medium of the scanned image.
The medium identifier may be an alphanumeric identifier and may have a hierarchical arrangement. Thus, particular sub-categories of medium may be identified by various levels of the hierarchical medium identifier.
In this way, the spectrum inversion apparatus and methods of this invention may accurately reproduce a scanned image by obtaining spectral space image values from a limited color space representation of the scanned image. Thus, the effects of metamerism may be reduced. These and other features and advantages of the invention will be described in greater detailed hereinbelow.